Smart Chronic Kidney Disease Detection and Progression

Monitoring System Using Machine Learning

Abhishek C, Dept of Computer Science and Engineering, Don Bosco Institute of Technology,
Bangalore, India [email protected]
Dinesh G, Dept of Computer Science and Engineering, Don Bosco Institute of Technology,
Bangalore, India [email protected]
Gowtham C, Dept of Computer Science and Engineering, Don Bosco Institute of Technology,
Bangalore, India [email protected]
Hitesh R C, Dept of Computer Science and Engineering, Don Bosco Institute of Technology,
Bangalore, India [email protected]

Corresponding Author (Guide)

Thejaswini D, Associate Prof Faculty, Dept of Computer Science and Engineering, Don Bosco
Institute of Technology, Bangalore, India

Email Address:
[email protected]

Abstract
Chronic Kidney Disease (CKD) is a chronic global health problem characterized by a gradual
loss of the functional capacity of the kidney, generally developing unchecked until a mature
level. Early detection is critical to avert disease advancement and reduce healthcare systems'
liabilities. This survey paper provides a comprehensive overview of emerging machine learning
(ML) methods developed for early prediction and disease progression monitoring of CKD. It
examines a wide variety of studies that use demographic, clinical, and biochemical data—
collected primarily from publicly available datasets such as the UCI Repository. The paper
outlines some of the ML models like Support Vector Machines (SVM), Random Forests (RF), k-
Nearest Neighbors (KNN), Decision Trees (DT), Multilayer Perceptrons (MLP), Gradient
Boosting, and ensemble learning techniques. These methods leverage feature selection, data
preprocessing, synthetic sampling (e.g., SMOTE), and hyperparameter optimization to enhance
model interpretability and robustness. In addition to early detection, there are some works aiming
at CKD stage classification and progression tracking with longitudinal data and real-time inputs.
This survey is contrasting the merits and demerits of different methods, addresses concerns such
as data imbalance and generalizability, and emphasizes the increasing emphasis on explainable
AI and clinical incorporation. The review concludes by establishing future research directions for
developing clinically deployable, scalable, and interpretable ML models focused on CKD
management.
Keywords
Chronic Kidney Disease, Machine Learning, Ensemble Methods, Feature Selection, Prediction
Models, Clinical Decision Support.
Introduction

Chronic Kidney Disease (CKD) is a severe and rising global health burden, defined by a
persistent reduction in kidney function over time. CKD is a progressive and irreversible
disease that usually undergoes five stages, ranging from mild kidney impairment to complete
renal failure. The asymptomatic nature of CKD at the early stage renders early diagnosis very
challenging, predisposing to delayed clinical presentation once the disease has advanced
significantly. Consequently, the capacity to make precise predictions and classify the stage of
CKD is significant in guiding treatment decisions, slowing progression, and reducing the risk
for complications such as cardiovascular disease and end-stage renal failure.

Recent machine learning (ML) developments have enabled the development of predictive
models that can learn complex paflerns from clinical information and offer novel
opportunities for non-invasive and early CKD diagnosis. These models utilize a combination
of features, including laboratory parameters (e.g., serum creatinine, albumin), patient
demographics, and glomerular filtration rate (GFR) to classify patients into appropriate
disease stages. Because CKD progression is nonlinear and influenced by a range of risk
factors, ML models can help clinicians provide uniform, data-driven assessments that can
improve upon standard diagnostic methods.

Several studies have contrasted different ML approaches to predict CKD stage, employing
classifiers such as Support Vector Machines (SVM), Random Forests (RF), and ensemble
methods. These techniques have been combined with stable preprocessing techniques such as
normalization, feature selection, and data balancing to handle challenges such as missing
values and class imbalance. Moreover, ensemble learning and real-time data fusion have been
discovered to increase the trustworthiness of models and their usability in clinical seflings.

The survey unites principal methodologies and findings of current studies committed to the
prediction of CKD stages through machine learning. The paper aims to be a synthesis of the
most frequent algorithms, feature engineering techniques, as well as evaluation metrics.
Besides, the article discusses the current limitations and indicates areas for further
development with particular focus on the development of interpretable, clinically feasible
models for enhancing the treatment of patients and decision-making in nephrology.

Despite increasing rates of CKD worldwide, the majority of affected cases go undiagnosed
until kidney damage is irreversible. Most traditional methods rely on fixed cut-offs and
periodic screening and thus may fail to capture the disease progress dynamics. In contrast,
machine learning models can continually execute on individual patient data in order to detect
paflerns of a very insidious, stage-wise deteriorative nature. These predictive capacities form a
helpful addition to clinical know-how, specifically in resource-constrained settings where
specialist access is limited.

Literature Review

Chronic Kidney Disease (CKD) is a chronic health disease that leads to progressive
kidney function loss and eventually to end-stage renal disease (ESRD) if not
diagnosed early. CKD is divided into five stages, based on glomerular filtration rate
(GFR), and detecting the stages early is crucial for early intervention. Machine
 learning (ML) algorithms have increasingly been used in recent years to forecast and
categorize CKD stages more accurately, enabling data-driven decision-making in
healthcare. This section offers a comprehensive literature review of major works in
this field.
Some studies have investigated the use of machine learning for CKD prediction and
management:
● Geetha et al. [1] (2025): in their risk prediction study of CKD, evaluated several machine
learning algorithms, including Random Forest (RF), Support Vector Machine (SVM), and
Multi-Layer Perceptron (MLP), alongside ensemble methods like hard and soft voting.
Their comparative study found that SVM achieved the maximum individual accuracy of
96.25%, and ensemble models demonstrated enhanced robustness against data variability.
The authors concluded that these ensemble-based models could be highly beneficial in
developing reliable decision-support tools for clinicians, aiding in the early and effective
management of chronic kidney disease. This research contributes to addressing CKD as a
growing global health issue through advanced predictive techniques.
● Pathak et al. [2] (2025): in their work on predicting chronic kidney disease, aimed to
enhance early and precise detection to prevent further health deterioration. They
emphasized the necessity of reliable prediction algorithms due to CKD's propensity to
advance to end-stage renal disease. Their study explored various machine learning
methods, such as Random Forest, k-Nearest Neighbors (KNN), Gradient Boosting, and
Naive Bayes, to improve CKD prediction. This research contributes to better management
of CKD by providing insights into effective predictive methodologies.
● Nandi et al. [3] (2025): in their paper on predictive modeling for Chronic Kidney Disease,
proposed a hybrid approach that integrates exploratory data analysis, machine learning,
and Explainable AI (XAI) techniques. The objective of this integration was to enhance
both the accuracy and the interpretability of CKD prediction models. This methodology
aims to provide more transparent and understandable predictions, which is crucial for
clinical decision-making in managing CKD. Their work highlights the benefits of
combining different analytical methods to improve the robustness and clarity of disease
prediction.
● Chantamit-o-pas et al. [4] (2025): in their study, evaluated the effectiveness of an
interactive robot-assisted avatar health education program for Chronic Kidney Disease
patients. They aimed to assess the program's impact on patients' knowledge and health
literacy, addressing the substantial health and economic challenges posed by CKD
globally. This research indicates that innovative educational tools can play a significant
role in empowering CKD patients with better understanding and self-management skills.
The findings suggest a promising avenue for improving patient engagement and health
outcomes in CKD care.Dhuriya et al. [5] (2025): in their study on Chronic Kidney
Disease, investigated data clustering and prediction using various machine learning
algorithms. They highlighted CKD as a major worldwide health concern with high
mortality and morbidity rates, often missed in early stages due to asymptomatic nature.
Their research aims to improve early detection and management of CKD by applying these
algorithms, contributing to more timely interventions. This work emphasizes the
importance of machine learning in addressing the global health crisis posed by CKD.
● Rani et al. [6] (2024in their study on XGBoost-driven insights for enhancing Chronic
 Kidney Disease detection, underscored the critical importance of early detection and
treatment to halt disease progression. They noted that recent advancements in machine
learning have significantly improved pattern detection and analysis in large datasets,
thereby enhancing CKD diagnosis and prediction. Their work leverages XGBoost to
achieve more accurate and efficient identification of CKD, contributing to better
management strategies. This research emphasizes the role of advanced ML techniques in
addressing CKD as a major global health issue.Chen et al. [7] (2024): Evaluated the
efficacy of a robot-assisted avatar education program for CKD patients, demonstrating
how new learning materials can empower patients.
● Bhuria [8] (2024): in their paper on advanced machine learning techniques for Chronic
Kidney Disease prediction and management, identified CKD as one of the top five deadly
diseases worldwide. The study introduced a highly accurate and sophisticated machine
learning approach for CKD prediction, achieving an accuracy rate of 97.5%. The model
demonstrated flawless precision and comprehensiveness (100%) for the non-CKD
category, ensuring reliable identification. This research highlights the potential of machine
learning to significantly improve early detection and enable timely interventions for
CKDSirisha et al. [9] (2024): Suggested a predictive model for predicting CKD stage
with Stochastic Gradient Descent (SGD) with ElasticNet regularization, with an
objective to perform effective classification without overfifling.
● Sonone [10] (2024): in their report on early prediction and progression of Chronic Kidney
Disease, described a groundbreaking study utilizing machine learning approaches with
real-time clinical datasets. The research aimed to develop predictive models for early CKD
detection and progression tracking, providing reliable insights into disease development.
By combining various clinical tests and patient data, the proposed method effectively
evaluates longitudinally gathered information. This study enhances the effectiveness of
machine learning algorithms for CKD by incorporating ensemble approaches, improving
accuracy and interpretability.
● Arora et al. [11] (2024): analyzed machine learning algorithms for Chronic Kidney
Disease prediction, emphasizing CKD as a critical global health concern. They highlighted
the urgency of early detection due to CKD's asymptomatic early stages and its significant
contribution to global mortality. Their study focused on developing a predictive model to
prevent End-Stage Renal Disease (ESRD) and facilitate efficient treatment. This research
contributes to improving diagnostic tools by evaluating various machine learning
algorithms for robust CKD predictionSneha [12] (2024): Formulated a supervised
predictive algorithm for CKD using ensemble methods such as AdaBoost, KNN, and
Random Forest that assisted in efficient early diagnosis.
● Sharma and Verma [13] (2024): in their analysis for predicting Chronic Kidney Disease
with machine learning models, identified CKD as a major global health crisis with a high
risk of progressing to end-stage renal disease. They stressed the vital importance of early
detection and diagnosis for prevention and timely treatment, as unhealthy dietary habits
and insufficient water consumption are significant causes. Their work applies machine
learning techniques like KNN, Decision Tree, and Random Forest to accurately predict
CKD. This study underscores the efficacy of these models in improving patient survival
through early intervention
● Makaraju and Mande [14] (2023): developed a Python-based machine learning model for
 Chronic Kidney Disease classification, recognizing CKD as a chronic illness with often
asymptomatic early stages. They highlighted the critical need for early detection and
treatment to prevent severe consequences arising from waste and fluid buildup. Their
model incorporates various input features, including baseline health measurements,
medication usage, medical history, and demographic data, to predict CKD risk. The
primary objective is to estimate CKD risk, outputting "1" for high-risk patients in stages 3
to 5.
● Anurag et al. [15] (2023): in their study on Chronic Kidney Disease prediction using a
robust machine learning approach, addressed CKD as a degenerative condition impairing
kidney function and causing waste buildup. They asserted that early CKD prediction and
identification can halt or delay the disease's progression. Their method combines feature
selection and classification algorithms using demographic, clinical, and lab data to
pinpoint key risk factors and provide accurate predictions. This approach aims to facilitate
effective therapies and improve outcomes for CKD patients.
● Kashyap et al. [16] (2022): in their study on predicting chronic disease in kidneys using
machine learning classifiers, aimed to overcome the challenges associated with late
diagnosis of CKD. They noted that advanced stages of CKD often leave patients with only
costly options like dialysis or kidney transplantation. Their research employed machine
learning classifiers to predict CKD, striving for early detection to enable timely
interventions and avoid severe consequences. This work emphasizes the potential of ML to
improve the management of this widespread health concern.

Comparative Analysis

Reference Methodology Models Used Key Features Limitations

[1] Ensemble model with RF, SVM, MLP, SVM accuracy: Ensemble (hard
RF, SVM, MLP Hard & Soft Voting 96.25%, ensemble voting) lower
robustness, real-time performance, no deep
intent learning used

[2] Model benchmarking RF, KNN, Gradient 100% accuracy in top Focused only on
with six algorithms Boosting, Naive models, 400 UCI classification; no
Bayes, CatBoost, samples XAI, real-world
AdaBoost deployment untested

[3] Exploratory analysis SVM, Decision Explainability, 17% Model validation

+ XAI (LIME) Trees, XAI (LIME) CKD prevalence inconsistencies, not
context, hybrid ML integrated clinically
models

[4] Chatbot deployment Dialogflow, Focus on stage 3 CKD; Only 50 patients;

using NLP models; Gemini, GPT-3.5 integration with LINE limited language
 clinical pilot with app; real-world support (Thai);
patients usability tested accuracy varies by
NLP model

[5] Clustering, Decision Tree, High precision No new dataset;

classification, Random Forest, (Random Forest 99%); relies on UCI CKD
imputation for Naïve Bayes, SVM, KNN imputation used dataset; lacks clinical
missing data KNN for missing data integration

[6] Feature selection, XGBoost High accuracy Limited dataset, lacks

XGBoost, Confusion (98.33%), F1-score generalization
matrix evaluation 98.63%, No false analysis
positives

[7] Quasi-experimental Robot-assisted Focus on health Small sample (n=30),

study with digital avatars, GEE literacy, 12-unit limited scalability
avatars analysis program, personalized
education

[8] Feature selection + Random Forest, 97.5% accuracy, hybrid Recall slightly lower
classification, hybrid Logistic techniques, no false (93.75%),
ML Regression, Extra positives generalizability not
Trees, SVM discussed

[9] SGD classifier with SGD, ElasticNet, Deployed via Streamlit, Lower accuracy than
ElasticNet Streamlit App feature selection, 85% other models, lacks
regularization accuracy clinical dataset
variability

[10] Longitudinal analysis Ensemble Learning, Tracks CKD Implementation

+ real-time dataset Clinical Test progression, real-time details sparse, lack of
Integration insights, biomarker use tool deployment

[11] Compared multiple Random Forest, Thorough comparative No details on real-

ML models on a SVM, Logistic evaluation of multiple time deployment;
dataset of 9993 Regression, KNN, algorithms; 100% lacks clinical trial
entries Naïve Bayes, accuracy with hybrid validation
Decision Tree models

[12] Supervised ML with Random Forest, Ensemble methods Dataset size limited
training/testing split AdaBoost, KNN used; practical clinical (400 entries); doesn't
implications discussed include external
validation

[13] Dataset from UCI Decision Tree, High accuracy Only benchmarked
CKD, preprocessing, Random Forest, (Random Forest 97%); traditional ML
ML model evaluation KNN emphasis on metrics models; lacks real-
like precision and F1- time system context
score
[14] Feature selection + ML (unspecified 89% accuracy, Accuracy lower than
classification mix), personalized top models, lacks
Demographics + predictions, intelligent deployment or app
Labs diagnosis layer

[15] Data preprocessing, Python, Logistic Utilized demographic, No mention of model

feature selection, Regression medical, medication, comparison; limited
binary classification and baseline health discussion of dataset
(0/1 for CKD stage) data; focuses on early performance
detection

[16] Comparative ML SVM, KNN, Preprocessing with Lacks model

model analysis using Random Forest, SMOTE, eGFR explainability, small-
UCI dataset Decision Tree staging, model scale testing
accuracy comparison

Results/Discussion
The literature reviewed repeatedly proves the ability of machine learning models to enhance the
early identification and progression tracking of Chronic Kidney Disease. The use of several ML
algorithms, such as SVM, Random Forests, KNN, Decision Trees, MLP, Gradient Boosting, and
ensemble approaches, has been implemented with success. All the performance measures,
including accuracy, precision, and recall, show great effectiveness in differentiating CKD stages
and detecting individuals at risk.
One of the most significant findings across various studies is the necessity of effective data
preprocessing methods such as normalization, feature selection, and dealing with missing values
and class imbalance (for example, applying SMOTE). These steps are essential for making
models robust and more generalizable. Ensemble learning techniques, especially, have been
found to be better performing and resistant to data variation than single models, emphasizing
their usefulness in sophisticated medical prediction problems.
The paper also emphasizes the increasing interest in explainable AI (XAI) methods. Models
providing interpretability, like those that incorporate LIME or SHAP, are becoming increasingly
sought after in clinical seflings because they allow for insights into the decision-making process,
promoting clinician trust. Real-time data integration and the creation of clinician-friendly tools,
like healthcare chatbots, are seen as major developments on the path to the practical usability of
such models.
Even with the promising outcomes, there are issues that persist. Data imbalance is a phenomenon
that is present quite commonly in CKD data sets and can cause biased models if it is not well
addressed. Another important area that needs more research is generalizability across a variety of
patient populations and clinical seflings. Closing the gap between algorithmic performance and
true clinical utility requires sustained work on scalable and deployable ML systems.
Future research should emphasis developing clinically validated ML systems that are able to
integrate smoothly into current healthcare workflows. This involves improving more advanced
feature engineering methods, examining more complex deep learning architectures, and
performing longitudinal studies to better comprehend disease progression. The ultimate aim is to
facilitate personalized care and proactive intervention for CKD patients, thus lessening the global
disease burden
Conclusion
Chronic Kidney Disease (CKD) represents a notable international health concern due to its
irreversible and typically undisclosed nature in initial stages. This survey has comprehensively
studied current developments within machine learning (ML) algorithm approaches for
prediction and monitoring in CKD stages early on. It underlined the efficacy of models like
Support Vector Machines, Random Forests, Gradient Boosting, and ensemble methods,
particularly when fortified with solid preprocessing, feature selection, and explainable AI
methodologies. Special aflention was given to the need for real-time integration of data,
interpretability of models, and deployment within clinical seflings. Although there has been
significant progress, there are still challenges in balancing data, being generalizable across
populations, and closing the algorithmic performance-utility gap. Future research will need to
aim at creating explainable, scalable, and clinically validated ML systems that can be easily
integrated into healthcare workflows to enable early diagnosis and personalized care for CKD
patients.

References
[1] Geetha M, Suganthe R C, Sowrabha A, & Sudharshan P. (2025). Chronic Kidney Disease Risk
Prediction Using Machine Learning. 2025 3rd International Conference on Disruptive Technologies
(ICDT), DOI: 10.1109/ICDT63985.2025.10986441
[2] Pathak, A., Singh, O. P., & Biswas, U. (2025). Predicting chronic kidney disease using machine
learning methodologies. 2025 8th International Conference on Electronics, Materials Engineering &
Nano-Technology (IEMENTech), DOI: 10.1109/IEMENTECH65115.2025.10959624
[3] Nandi, A., Ghosh, A., Paul, P., Singh, A., Das, A. K., & Gupta, A. (2025). Predictive Modelling
for Chronic Kidney Disease: A Hybrid Approach Using Exploratory Data Analysis, Machine
Learning and Explainable AI. 2025 8th International Conference on Electronics, Materials
Engineering&Nano-Technology(IEMENTech),DOI: 10.1109/IEMENTECH65115.2025.10959624
[4] Chantamit-o-pas, P., Kulsuwan, S., Suflhiwichienchot, N., & Ponpinij, P. (2025). Healthcare
Chatbot for slowing the Progression in Chronic Kidney Disease Stage 3 patients. 2025 International
Conference on Artificial Intelligence in Information and Communication (ICAIIC),
DOI: 10.1109/ICAIIC64266.2025.10920642
[5] Dhuriya, G., Yadav, N., Saini, A., & Afliuflama. (2025). Chronic Kidney Disease: Data Clustering
and Prediction Using Various Machine Learning Algorithms. 2025 3rd International Conference on
Disruptive Technologies (ICDT), DOI: 10.1109/ICDT63985.2025.10986563
[6] Rani, R., Gill, K. S., Upadhyay, D., & Devliyal, S. (2024). XGBoost-Driven Insights: Enhancing
Chronic Kidney Disease Detection. 2024 5th International Conference on Smart Electronics and
Communication (ICOSEC), DOI: 10.1109/ICOSEC61587.2024.10722440
[7] Chen, N.-J., Chang, C.-H., Lai, C.-L., Liu, K.-Y., & Guo, J.-L. (2024). Effectiveness of an
Interactive Robot-Assisted Avatar Health Education Program on Knowledge and Health Literacy in
Chronic Kidney Disease Patients. 2024 IEEE International Conference on E-health Networking,
Application & Services (HealthCom), DOI: 10.1109/HEALTHCOM60970.2024.10880787
[8] Bhuria, R. (2024). Advanced Machine Learning Techniques for Chronic Kidney Disease
Prediction and Management. 2024 International Conference on Artificial Intelligence and Emerging
Technology (Global AI Summit), DOI: 10.1109/GLOBALAISUMMIT62156.2024.10947932
[9] Sirisha, A. B. N. P., Dhana Lakshmi, N., & Siva Priyanka, S. (2024). Chronic Kidney Disease Stage
Prediction Using SGD with Elastic Net Regularization: A Streamlit Application. 2024 International
Conference on Computing, Sciences and Communications (ICCSC), DOI:
10.1109/ICCSC62048.2024.10830323
[10] Sonone, N. (2024). EARLY PREDICTION AND PROGRSSION OF CHRONIC KIDNEY
DISEASE USING MACHINE LERNING TECHNIQUES. 2024 2nd International Conference on
Networking and Communications (ICNWC), 1 DOI: 10.1109/ICNWC60771.2024.10537571
[11] Arora, A., Sehgal, C., & Agarwal, N. (2024). An Analysis of Machine Learning Algorithms for
Chronic Kidney Disease Prediction. 2024 14th International Conference on Cloud Computing, Data
Science & Engineering (Confluence), DOI: 10.1109/CONFLUENCE60223.2024.10463505
[12] Sneha, A. (2024). Development of Supervised Algorithm for Predicting Chronic Kidney
Disease. 2024 IEEE International Conference on Computing, Power and Communication
Technologies (IC2PCT), DOI: 10.1109/IC2PCT60090.2024.10486505
[13] Sharma, T., & Verma, N. (2024). Analysis for Predicting Chronic Kidney Disease with the
Application of Machine Learning Models. 2024 5th International Conference on Data Intelligence and
Cognitive Informatics (ICDICI), DOI: 10.1109/ICDICI62993.2024.10810875
[14] Makaraju, K. L., & Mande, N. V. S. (2023). Machine Learning Predictive Model for Chronic
Kidney Disease Classification Using Python. 2023 International Conference on Artificial Intelligence
for Innovations in Healthcare Industries (ICАПНІ),
DOI:10.1109/ICAIIHI57871.2023.10489064
[15] Anurag, Vyas, N., & Balla, D. (2023). Chronic Kidney Disease Prediction Using Robust
Approach in Machine Learning. 2023 3rd International Conference on Innovative Sustainable
Computational Technologies (CISCT), DOI: 10.1109/CISCT57197.2023.10351277
[16] Kashyap, C. P., Reddy, G. S. D., & M, B. (2022). PREDICTION OF CHRONIC DISEASE IN
KIDNEYS USING MACHINE LEARNING CLASSIFIERS. 2022 1st International Conference on
Computational Science and Technology (ICCST), 2 DOI:10.1109/ICCST55948.2022.10040329
Figures and Tables

High CKD Dat Preprocess

Urinary Disbeties Blood Raw a ed
Tract Presure
Infections Real Preproc data
(UTI) Time
Age

Swelling CHRONIC
KIDNEY Multi Performance Evaluation
DISEASE Family Classifier of
Changes in (CKD) History Algorithm Algorithms
Urine outpt
(Foamy
Urine)
Medications
High (Nonsteroidal
Smoking
Cholesterol anti- CKD With NOT
inflammatory
Drugs Progression CKD
Stage

Figure 1: Common Factors Cause Chronic Kidney Disease Figure 2: Proposed Methodology

CKD Stages & eGFR Medical Interventions

Range
1st Stage - Over 90 ml per Yearly check up on kidney
minute per 1.73 m² function.
2nd Stage - between 60 Checkup of kidney
to 89 ml per minute per function regularly
1.73 m²
treatment of underlying
diseases like diabetes and
hypertension.
3rd Stage – Within the Regular checkup of kidney
range of 30 to 59 ml per function
min per
Regulate problems, such as
1.73 m² anemia and bone disease.
4th Stage - Between15 Regulate blood pressure and
and 29 ml per min per blood sugar levels
1.73 m²
Regulate complications (e.g.,
anemia, bone disease)
Consider referring the patient
to nephrologist
5th Stage - Below 15 ml Initiate renal replacement
per min per 1.73 m². therapy (transplant or
dialysis)
Prepare the patient for renal
replacement therapy.
Educate the patient on
therapy options

Table 1: Various Stage of Chronic Kidney Disease and Medical Interventions